Find the cheapest API for any model and connect it from your AI agent: Grandet Skill and an anonymous MCP server, backed by listed prices from 1,800+ official, inference and relay providers on grandet.ai.

Grandet – Cheapest‑API Finder for LLMs

What it is – Grandet is a lightweight, open‑source “skill” that lets an AI‑assistant (Claude Code, Cursor, Codex, etc.) look up the lowest‑priced endpoint for a given large‑language‑model. It pulls price data from official providers, third‑party inference services and relay sites, and can also surface free‑credit offers.

Why it matters – When building an app that calls LLM APIs, the cost can dominate the budget. Grandet automates the tedious research step, letting developers or agents pick the cheapest usable endpoint without manually scanning dozens of provider pages.

Core functionality

Feature	What it does
Cheapest‑price lookup	Queries Grandet’s catalogue (hosted at `grandet.ai`) for a model and usage type (chat, coding, long‑output) and returns the few lowest listed prices, the site, token‑group, unit price and the observation date.
Free‑credit discovery	Lists official free tiers, trial credits, and promotional offers (e.g., sign‑up bonuses) together with any required verification steps.
One‑click connection	After you pick a provider, Grandet generates the exact configuration command for the target IDE/agent, backs up the existing file, and (after your confirmation) writes the new settings. The secret key never passes through the chat interface.
MCP server	Exposes the same lookup logic over a streamable HTTP endpoint (`https://grandet.ai/mcp`) that any MCP‑compatible client can call anonymously, without an API key.
Glossary & evidence	Provides a built‑in glossary of relay‑jargon and links to the original price sources for transparency.

How to use it

Install the skill – In the chat window of a supported agent type:
```
Install Grandet Skill for me: https://grandet.ai/download.md
```
The agent follows the link, downloads the skill files and registers the new commands.
Ask for a price – Example:
```
Find me the cheapest Claude Sonnet 4.6 API.
```
Grandet replies with a short table of the cheapest options and any free‑credit notes.
Connect – Say something like:
```
Connect the second one to Codex on Windows.
```
Grandet walks you through signing up, creates the token, and then writes the appropriate configuration line (or gives you the exact CLI command) after you confirm.
Undo – If you change your mind, say “undo” and Grandet restores the previous config.

Installation for developers

Prerequisite: Node 20+ (no external npm packages required).

Typical workflow:

# Validate the catalogue files
node scripts/catalogue.mjs validate
# Regenerate derived data (optional)
node scripts/catalogue.mjs generate --check
# Run the local test suite
node --test tests/*.test.mjs
# Serve a local preview of the site
node scripts/serve.mjs   # opens http://127.0.0.1:8787/

Manual skill install – See INSTALL.md for copying the skills/grandet‑api‑price‑finder folder into the target agent’s skill directory or uploading the download.md file.
MCP client – Add to any MCP‑compatible client with a JSON snippet, e.g.:
```
{ "mcpServers": { "grandet": { "url": "https://grandet.ai/mcp" } } }
```

When it shines

Cost‑sensitive prototypes – Quickly compare dozens of providers before committing to a paid plan.
Exploratory research – Discover free trial credits or promotional offers you might otherwise miss.
Multi‑agent environments – The same skill works across Claude Code, Cursor, Gemini CLI, Hermes, etc., providing a uniform interface.

Limitations

Grandet only reports listed prices; it does not evaluate latency, reliability, or model fidelity.
Prices are shown in the currency they were advertised (USD, CNY, cash) and are not automatically converted.
The “connect” step still requires you to create an account and obtain a token on the chosen provider’s site.
Data is as current as the last catalogue release; real‑time price changes are not reflected until the site is updated.

Licensing

Skill files – Apache‑2.0.
Catalogue code & documentation – AGPL‑3.0‑only (see LICENSE).
Supplier data – Not under an open‑data licence; it remains the property of the original providers (see DATA‑LICENSE.md).

Quick links

Website / price UI – https://grandet.ai
MCP endpoint – https://grandet.ai/mcp
API description – https://grandet.ai/public-data/price-finder.json
Free‑credit catalogue (JSON) – https://grandet.ai/data/free-tokens-v3.json
Contributing – See CONTRIBUTING.md and METHODOLOGY.md.

Grandet lets you let your AI agents do the price‑shopping, so you can focus on building the product.

hanzohanzhe/Grandet